Streamsight

Streamsight is an offline Reccomender Systems (RecSys) evaluation toolkit that respects a global timeline. The aim is to partition the data into different windows where data is incrementally released for the programmer to fit, train and submit predictions. This aims to provide a close simulation of an online setting when evaluating RecSys algorithms. This library is built on top of the original V1 Streamsight.

Full Flow Structure

Getting Started

1. Clone the repository

git clone https://github.com/suenalaba/streamsightv2
cd streamsightv2

2. Install dependencies locally

python3 -m venv venv
source venv/bin/activate
pip install .

Alternatively, dependencies can be installed with poetry

pip install poetry
poetry install

The dependencies are listed in pyproject.toml.

Contributing

• We welcome all contributors, be it reporting an issue, or raising a pull request to fix an issue.
• When you make changes, rerun pip install . to test your changes.

Documentation

The documentation can be found here and repository on Github.

Publishing

1. Run the following command to build the library

poetry build

2. Ensure that your config has been set with

poetry config pypi-token.pypi <YOUR_PYPI_API_TOKEN_HERE>

3. Publish the package

poetry publish

Citation

If you use this library in any part of your work, please cite the following papers:

Ng, T. K. (2024). Streamsight: a toolkit for offline evaluation of recommender systems. Final Year Project (FYP), Nanyang Technological University, Singapore. https://hdl.handle.net/10356/181114